Review of John Galt Solutions, Supply Chain Planning Software Vendor

By Léon Levinas-Ménard

Last updated: November, 2025

John Galt Solutions is a US-based supply chain planning software vendor founded in 1996 that has evolved from an early focus on Excel-based forecasting (ForecastX) to a cloud-based, AI-branded Atlas Planning Platform covering demand planning, S&OP/IBP, inventory and supply planning, production scheduling and related workflows for mid-market and larger enterprises. It positions Atlas as a single SaaS platform on Microsoft Azure that “connects and orchestrates” end-to-end planning, with recent emphasis on probabilistic planning, multi-echelon inventory optimization (MEIO), ensemble forecasting, and explainable AI features built on top of machine learning models. The company appears privately held and largely bootstrapped, with no public venture rounds, and markets a consulting-heavy deployment model working with named customers such as Reddy Ice, Mars, Netgear, Ping, Sara Lee Frozen Bakery and others across CPG, food and beverage, aftermarket parts, industrial manufacturing and retail. Publicly available documentation, analyst notes and press releases confirm commercial traction and broad functional coverage, but provide limited hard detail on the underlying algorithms, data structures or engineering stack; as a result, many of the “AI” and “advanced optimization” claims remain marketing-level and need to be interpreted conservatively when assessing how state-of-the-art the technology truly is.

John Galt Solutions overview

From a corporate perspective, John Galt Solutions (JGS) is an independent software vendor focused on forecasting and supply chain planning. English-language sources describe the firm as founded in 1996 by Anne Omrod (often spelled Omrod/Omrod) and Mark Holm, initially headquartered in Chicago, Illinois.¹² Japanese-language coverage of the company, which cross-references early technical publications, similarly dates its origin to the mid-1990s and emphasizes its roots in business forecasting and data warehousing.³ The company’s own “Our Story” page simply states that it has been “partnering with leading companies” since 1996 without disclosing ownership or funding structure.⁴ Third-party profiles (CB Insights, Tracxn, various vendor directories) list John Galt as privately held with no major institutional rounds recorded, suggesting organic or lightly angel-funded growth rather than classic VC scaling.⁵⁶⁷

Functionally, JGS presents itself as a specialist in demand forecasting and end-to-end supply chain planning. The public website leads with “The fastest way to get more value from your supply chain,” and offers two main products: the Atlas Planning Platform (a cloud SaaS platform) and ForecastX (an Excel add-in for one-click forecasting).⁸ Analyst reports and vendor directories consistently describe Atlas as covering demand planning, S&OP/IBP, inventory optimization, supply planning, production scheduling and distribution/delivery planning for mid-market and enterprise clients across multiple industries.⁵⁹¹⁰ ForecastX, by contrast, is aimed at planners who remain primarily in spreadsheets, providing time-series forecasting methods via Excel.⁸

Over roughly the last decade, JGS has shifted its branding from “forecasting and demand planning tools” toward “AI-driven end-to-end supply chain planning” with Atlas positioned as its flagship offering and ForecastX as a lighter-weight entry point.³⁸ Atlas is available in the Microsoft Azure Marketplace, where Microsoft describes it as a “single, AI-driven Software as a Service (SaaS) solution” that connects and orchestrates the end-to-end supply chain, spanning S&OP/IBP, demand, supply, inventory, production and delivery planning.⁹ Recent press releases emphasize advanced analytics, probabilistic planning, multi-echelon inventory optimization (MEIO), ensemble forecasting, and more recently explainable AI (xAI) and generative AI assistants (“Galt Intelligence”) layered on top of Atlas.¹¹

Commercially, JGS claims a broad roster of customers including well-known brands like Mars, Netgear, Ping, Sara Lee Frozen Bakery, Reddy Ice and others, showcased via logos and case studies on the corporate site.⁸ Independent case-study aggregators confirm at least some of these relationships (e.g., Reddy Ice).¹² Analyst firms such as Nucleus Research and Gartner consistently include JGS/Atlas in their supply chain planning vendor landscapes, sometimes highlighting usability and time-to-value rather than deep technical innovation as the primary differentiators.¹³¹⁴ Overall, the picture is of a mature, commercially established vendor with significant domain experience, whose public technical narrative is heavy on AI/automation language but relatively light on verifiable, low-level engineering details.

Company background and history

Early descriptions of JGS frame it as a forecasting and data-warehousing specialist. A widely-cited support article (“Who is John Galt?”) describes the firm as “a private company, founded in 1996 and focused on forecasting and demand planning solutions.”² Japanese Wikipedia notes that in the late 1990s the company developed an Excel-based forecasting wizard that performed strongly in the M3 forecasting competition, and that its founders co-authored the textbook Business Forecasting: Practical Problems and Solutions—placing JGS firmly in the classical statistical forecasting tradition rather than as a pure software start-up.³

Across various third-party profiles (Serchen, Technology Evaluation, CB Insights, Tracxn), JGS is consistently categorized as a mid-sized supply chain planning vendor, with estimated employee counts ranging from tens to low hundreds and headquarters listed in Austin, Texas in more recent records.³⁵¹⁵ A 2025 Newswire press release announcing a webinar with Reddy Ice lists the company’s contact address as “5900 Balcones Drive Ste 4629, Austin, TX 78731,” corroborating the shift of headquarters to Austin.¹⁶ No evidence of major acquisitions (either as acquirer or acquired) could be found in press releases, news databases or corporate filings; recent press communications focus on product enhancements, analyst recognitions and customer stories rather than M&A activity.¹¹¹⁶¹³

On funding, CB Insights and similar databases list John Galt Solutions with no disclosed VC rounds, labeling it effectively as “non-VC-backed.”⁶ Tracxn categorizes the firm within “Supply Chain Tech / Supply Chain Planning” and likewise shows no institutional fundraising events, suggesting that growth has been primarily customer-funded.⁷ There are no public indications of IPO plans or private equity transactions as of late 2025.

Product evolution

Historically, JGS’ first widely recognized product was ForecastX, an Excel add-in providing forecasting methods such as exponential smoothing, Box-Jenkins, and other time-series techniques within a spreadsheet interface. Japanese sources explicitly link ForecastX to the company’s early success, including strong performance in forecasting competitions.³ ForecastX remains actively marketed today as “powerful one-click forecasting in Microsoft Excel,” targeted at organizations that are “still spreadsheet-driven” but want better statistical forecasts and collaboration within that environment.⁸

The Atlas Planning Platform is the newer, strategic product line. Technology Evaluation and other analysts describe Atlas as a unified supply chain planning suite covering at least six process areas: S&OP/IBP, demand planning, inventory planning, supply planning, production scheduling and distribution/delivery (logistics) planning.⁵⁹ The Azure Marketplace listing, which is written from Microsoft’s perspective, reinforces this, describing Atlas as a “SaaS solution to connect and orchestrate the end-to-end supply chain” with focused modules for S&OP/IBP, demand planning (continuous planning to sense and shape demand), supply planning (synchronizing resources and constraints), inventory planning (raising service levels while freeing working capital), deliver (converting orders into “intelligent shipments”) and schedule (factory resource and capacity optimization).⁹

Over the last several years, JGS has added marketing emphasis on AI, probabilistic planning and multi-echelon inventory optimization. The main site highlights “Advanced Analytics,” “Probabilistic Planning,” “Artificial Intelligence,” and “What-if Scenarios” as core platform capabilities.⁸ A 2025 press release announces expanded explainable AI features, positioning Atlas as combining “advanced analytics with explainable AI (xAI)” and leveraging generative AI (GenAI) to increase transparency and user trust in MEIO and ensemble forecasting.¹¹ The same release references a GenAI assistant (“Galt Intelligence”) embedded in Atlas to explain MEIO recommendations and ensemble forecast results conversationally.¹¹ These additions suggest an incremental evolution: from spreadsheet forecasting (ForecastX), to a multi-module planning suite (Atlas), and more recently to layering AI/xAI features over that suite to improve usability and trust.

John Galt Solutions vs Lokad

Both John Galt Solutions and Lokad focus on supply chain planning and inventory decisions, but they approach the problem with notably different technical philosophies and delivery models. JGS offers Atlas as a configurable, user-facing SaaS application suite organized around classical planning processes—demand planning, S&OP/IBP, inventory planning, supply planning, scheduling—with workflows, UI configuration and scenario analysis as the primary levers for adaptation.⁸⁹ Lokad, by contrast, positions its platform as a programmable environment for “predictive optimization of supply chains,” where virtually all logic (data preparation, forecasting, optimization and economic evaluation) is expressed in a domain-specific language (Envision) and executed on a custom distributed engine.¹⁷ In other words, JGS sells an application; Lokad sells a coding-centric platform plus a team of “supply chain scientists” to build custom optimization apps on top.

On the forecasting side, JGS’ public collateral references “probabilistic planning,” “ensemble forecasting” and the use of AI/ML models across demand and inventory processes but does not document the underlying statistical architecture in detail.⁸⁹¹¹ There is no public technical documentation that spells out, for example, whether Atlas produces full demand distributions, how lead-time uncertainty is modeled, or how ensemble methods are constructed and calibrated. Lokad, on the other hand, explicitly centers its product on probabilistic forecasting; its technical documentation describes an engine that produces integrated demand distributions over lead time and exposes probabilistic random variables directly within its DSL, allowing computations over full distributions rather than single point forecasts.¹⁸ Lokad’s public materials also detail how those distributions are used for decision-centric optimization (e.g., prioritised inventory replenishment with economic drivers).¹⁸¹⁹ As a result, while both vendors advertise “probabilistic” capabilities, Lokad offers substantially more transparent and granular information about how such forecasts are computed and exploited.

In optimization and “AI,” the divergence continues. John Galt markets Atlas as an AI-driven platform with capabilities such as multi-echelon inventory optimization, simultaneous multi-objective optimization, and a GenAI-powered xAI layer (“Galt Intelligence”) that explains MEIO and ensemble decisions to planners.¹¹²⁰ However, the specific optimization algorithms (e.g., whether they use mixed-integer programming, heuristics, or simulation-based search) are not disclosed, and there is no public code or technical paper clarifying how MEIO or multi-objective trade-offs are solved numerically. By contrast, Lokad frames its technology stack explicitly around probabilistic forecasts plus custom stochastic optimization methods (e.g., Stochastic Discrete Descent) and highlights the use of differentiable programming to jointly learn forecasting and decision parameters.¹⁷¹⁹ Although Lokad’s algorithms are also proprietary, the company publishes technical documentation and lectures that explain the numerical structure of its optimization (randomized search over discrete decisions evaluated against Monte-Carlo scenarios, objective functions expressed in economic terms, etc.), offering more technical depth than typical enterprise marketing.

From a user-experience standpoint, Atlas is designed as a more classical enterprise application: planners work in browser-based screens to manage hierarchies, run plans, adjust assumptions, and analyze scenarios through dashboards and predefined workflows. Independent reviews frequently praise Atlas for usability and the ability to configure the UI and business processes without “lengthy custom builds,” highlighting relatively quick implementations (e.g., 3–6 months) and planning-team ownership of processes.¹⁰²¹ Lokad’s environment is closer to a programming studio: core artifacts are Envision scripts and data tables, and planners typically consume the output as prioritized decision lists or dashboards orchestrated by Lokad’s supply chain scientists.¹⁷¹⁹ This makes Lokad more flexible for modeling idiosyncratic constraints and economic drivers, but at the cost of higher technical involvement; JGS trades some flexibility for a more conventional, planner-friendly application.

Finally, there is a philosophical difference in how decisions are framed. JGS’ messaging emphasizes “orchestrating the end-to-end supply chain,” aligning planning processes and providing planners with AI-enhanced recommendations and explanations, but mostly within standard S&OP/IBP and planning constructs.⁸⁹¹¹ Lokad explicitly puts financial outcomes and probabilistic risk at the center, advocating for a decision-centric view where every recommendation is evaluated in monetary terms (e.g., expected profit or cost) and presented as a prioritized action list, rather than as a plan tied to fixed service-level KPIs.¹⁷¹⁹ For organizations whose culture is strongly process-driven and S&OP-centric, Atlas’ structure may be more familiar; for those seeking a more radical, economics-first re-design of planning, Lokad’s approach is structurally different.

Technology and architecture

Platform architecture and deployment model

Atlas is delivered as a multi-tenant SaaS application hosted on Microsoft Azure, as evidenced by its availability in the Azure Marketplace and Microsoft’s description of it as a “single, AI-driven Software as a Service (SaaS) solution” for end-to-end supply chain planning.⁹ The marketplace listing indicates a standard cloud deployment model: customers subscribe to Atlas through Azure, and the platform provides planning capabilities across multiple process areas from a single, integrated environment.⁹ There is no public evidence that on-premise deployments are still a mainstream option; historical materials suggest Atlas originated as on-prem/hosted but has been repositioned as cloud-native over time.

The public website implicitly confirms a unified platform architecture by presenting Atlas as “one platform for all your planning needs,” with separate application tiles (Demand, S&OP/IBP, Inventory, Supply, Deliver, Schedule) sitting on top of shared platform capabilities like Advanced Analytics, Probabilistic Planning, What-if Scenarios, Artificial Intelligence, Socialization and Sustainability.⁸ Marketing copy stresses a “complete view across supply and demand,” the ability to replace spreadsheets and siloed legacy tech, and “configurable” processes without “lengthy custom builds.”⁸ Customer testimonials quoted on the homepage describe consolidating multiple data sources into Atlas as a “single source for all data” within a few months of implementation.⁸

However, beyond these high-level statements, there is no public technical documentation of Atlas’ internal architecture (e.g., whether it uses microservices, what databases or message buses are involved, or how data models are structured). References to “digital supply chain transformation” and “configuration” suggest a relatively standard modern enterprise SaaS stack: a web UI, a central data model, and an application server tier exposing planning functions as services. Analyst writeups (e.g., Nucleus Research’s Value Matrix) echo this by emphasizing usability and time-to-value rather than novel architectural patterns.¹⁴ In the absence of code or detailed diagrams, one must assume Atlas follows typical SaaS design norms rather than any documented, radical architectural innovation.

Technology stack and interfaces

Third-party technology-profiling sites (e.g., Enlyft, similarweb-style tools) list JGS as using mainstream web technologies (JavaScript frameworks, marketing/analytics tags) but do not expose the internal server-side stack; there is no official statement that Atlas is built on a specific language or database.⁵ A developer-facing link to “Developer APIs” on the main site points to a Zendesk-hosted documentation portal, indicating that Atlas exposes APIs for integration and potentially custom applications, but the documentation itself is gated and not publicly indexable from the outside.¹⁶

Integration is marketed through “Galt Connect,” which is listed among platform capabilities on the main site and described in collateral as an integration framework connecting Atlas to ERPs, CRMs, WMSs, and external data sources such as POS or weather feeds.⁸⁹ A partnership announcement with enVista (a consulting and technology services firm) frames Atlas as part of a broader Azure-based ecosystem for supply chain and distribution planning, with enVista providing integration and implementation services—further reinforcing that integration relies on typical API/connector patterns rather than bespoke on-prem components.²²

Given the lack of public low-level information, any deeper claims about the internal technology stack (languages, frameworks, database technologies) would be speculative. What can be said with confidence is that Atlas is delivered as a browser-based SaaS platform with APIs, an integration layer (Galt Connect), and a multi-module planning UI built to be configured rather than coded by end-users.⁸⁹

AI, machine learning and optimization capabilities

Atlas’ marketing strongly emphasizes AI and advanced analytics. The Azure Marketplace listing explicitly notes that Atlas “brings a rich history of innovation in supply chain planning, advances in machine learning and AI” and that it supports real-time continuous planning in demand, as well as intelligent inventory and supply planning.⁹ The main website highlights “Artificial Intelligence” and “Probabilistic Planning” among core capabilities, without detailing the models used.⁸

Recent press releases are more specific about where AI is applied, though not how. A September 2025 press release announces that Atlas “expands explainable AI to build trust in supply chain decisions,” stating that new xAI features apply generative AI to provide transparency and context in multi-echelon inventory optimization (MEIO) and ensemble forecasting.¹¹ According to this release, Atlas now provides explanations for MEIO recommendations (e.g., where and why inventory changes are recommended, risk-pooling opportunities) and for ensemble forecasting (e.g., why particular models or patterns were selected in an ensemble forecast), via a conversational assistant called “Galt Intelligence.”¹¹ The same press release portrays MEIO and ensemble forecasting as pre-existing capabilities whose adoption had been hampered by perceived black-box behavior; xAI is presented as a usability and trust layer above those algorithms.

A separate release (not reproduced here due to space) discusses “enhanced simultaneous multi-objective optimization,” suggesting that the Atlas optimization engine can consider multiple objectives (e.g., service, cost, sustainability) in a single model.²⁰ However, no details are provided regarding the underlying solver technology (e.g., whether these are linear/quadratic programs, metaheuristics, or scenario-based search), nor is there any external, independent technical validation of the optimization algorithms’ performance.

Critically, there is no public technical documentation of the forecasting engine comparable to, say, an open whitepaper on model classes, error metrics or training procedures. References to “ensemble forecasting” imply that multiple models are combined (as is common in modern forecasting), and references to “probabilistic planning” imply that at least some outputs are distributions rather than point estimates, but these implications remain marketing-level without code or detailed documentation.⁸⁹¹¹ Independent reviews on SoftwareAdvice and G2 focus on user experience—configuration flexibility, dashboarding, planning workflows—and do not shed light on algorithmic details.¹⁰²¹

The safest interpretation, based on available evidence, is that Atlas does indeed incorporate machine learning models (likely a mix of time-series methods and more modern ML), uses ensemble techniques for forecasting, and provides some multi-objective optimization capability for inventory and supply planning. The “AI” and “xAI” branding is primarily applied to how these models’ outputs are exposed to users (e.g., via explanations, scenario analysis, conversational assistants), rather than to a documented, state-of-the-art algorithmic breakthrough.

Data, scenarios and workflow

The Atlas homepage and Azure listing both emphasize the ability to pull together data from multiple ERPs into a unified planning view, run scenarios, and support cross-functional S&OP/IBP alignment.⁸⁹ Customer quotes describe integrating “multiple ERPs into Atlas to obtain supply chain visibility and take action across multiple business units in half a year,” and going from “multiple data sources in Excel to one source for all data.”⁸ The Azure listing describes “real-time continuous planning to sense, shape and satisfy demand,” and scenario-based S&OP that links tactical and strategic horizons.⁹

Scenario analysis is explicitly listed as a platform capability (“What-if Scenarios”), and Atlas’ application pages (not detailed here) show UI concepts for adjusting assumptions, running alternative demand or supply scenarios, and comparing outcomes.⁸ This is consistent with mainstream supply chain planning tools: user-driven scenarios with pre-configured levers and outputs, rather than free-form probabilistic analysis.

The multi-echelon inventory optimization and ensemble forecasting features referenced in the explainable AI press release suggest that Atlas maintains a multi-tier network model and uses some form of probabilistic modeling to propagate demand and inventory risk across the network.¹¹ However, in the absence of technical documentation, it is not possible to determine how sophisticated these models are (e.g., whether they fully account for correlated uncertainties, stochastic lead times, or complex BOM structures) or how computationally intensive the optimization is at industrial scale.

Implementation and commercial maturity

Implementation approach

JGS positions itself as a partner that “works with you every step of the way” and offers services for digital transformation, implementation and post-implementation support, as well as an “Innovation Lab” and training/certification programs.⁸ The main site describes Atlas as “easy to configure – no lengthy custom builds required,” and customer testimonials mention implementation timelines on the order of three months for some projects.⁸ These claims align with feedback on review platforms where users often highlight relatively straightforward implementation and the vendor’s willingness to tailor configurations.¹⁰²¹

The presence of an integration-oriented capability (Galt Connect) and partnerships with consultancies like enVista indicate that JGS often operates with a consulting or SI partner to connect Atlas to customer ERPs, WMSs and other systems.²² Live webinars with customers (e.g., Reddy Ice) showcase how Atlas is used to integrate weather data, POS data, IoT sensor data and driver handheld data to drive automated demand and replenishment planning, suggesting that implementations can incorporate a fairly rich set of external signals when customers are ready to invest in such integrations.¹⁶

Overall, implementation appears to follow standard enterprise SaaS patterns: data extraction from existing systems, configuration of planning hierarchies and workflows, iterative tuning of models and parameters, and gradual adoption of recommendations into operational processes. There is no public evidence of formal, code-heavy customization (e.g., customer-written extensions), reinforcing that the platform is configured rather than programmed by clients.

Named customers and sectors

JGS publicly lists a broad range of industries targeted by Atlas: Apparel & Footwear, Aftermarket Parts, Beverages/Wines/Spirits, Chemicals, Consumer Durables, Consumer Products, Food & Nutrition, Hi-Tech & Electronics, Industrial Manufacturing, Life Sciences, Retail and Wholesale Distribution.⁸ Customer logos on the homepage include Amcor, Mars, Netgear, Ping, Sara Lee Frozen Bakery, and others.⁸

Case studies provide more concrete evidence. A FeaturedCustomers case study details how Reddy Ice used Atlas to improve service levels and reduce out-of-stocks across its distributed ice manufacturing and distribution network.¹² A 2025 Newswire press release describes Reddy Ice as “the world’s largest manufacturer and distributor of packaged ice” and explicitly calls it a “valued customer of John Galt Solutions,” highlighting a webinar on how Reddy Ice uses advanced planning technology (Atlas) with data from weather, POS, IoT sensors, and driver handheld devices for highly automated, agile planning.¹⁶ Other case-study thumbnails on the JGS site mention PING (golf equipment, with a 24-month rolling forecast model), Valent BioSciences (moving from spreadsheets to Atlas to reduce inventory and stockouts), and Mars (global digital supply chain transformation across 60 countries).⁸

While the exact scope and depth of each deployment are not fully disclosed, these public references, combined with analyst coverage, provide reasonable evidence that JGS has a diverse, international customer base using Atlas in production for at least demand and supply planning, with some customers also adopting more advanced optimization (e.g., MEIO, scenario-driven S&OP).

Analyst coverage and market position

Analyst firms consistently include JGS/Atlas in their supply chain planning landscapes. Nucleus Research’s 2022 and subsequent Supply Chain Planning Technology Value Matrix reports place Atlas in the “Leader” quadrant, often highlighting usability, time-to-value and customer satisfaction as key strengths.¹⁴ Gartner’s Magic Quadrant for Supply Chain Planning Solutions mentions John Galt Solutions as a vendor in the SCP space (positions such as “Challenger” or “Niche Player” vary across years, and precise quadrant positions are behind paywalls), indicating that Atlas is recognized as a credible option in the global SCP market.¹³

Review platforms like SoftwareAdvice and G2 aggregate user ratings for Atlas Planning Suite, generally reflecting high satisfaction with ease-of-use, vendor responsiveness and flexibility, but providing limited insight into the depth of AI/optimization capabilities beyond what JGS itself claims.¹⁰²¹ Taken together, this suggests that JGS is a commercially mature vendor with a solid reputation in mid-market and some enterprise segments, recognized more for practical usability and customer service than for radically novel technical architecture.

Conclusion

Based on available public information, John Galt Solutions delivers a commercially mature, broadly functional supply chain planning platform (Atlas) and a legacy-yet-still-used Excel forecasting tool (ForecastX). The firm has existed since the mid-1990s, appears to be privately held and organically grown, and has demonstrable traction with recognizable brands across multiple industries. Its inclusion in analyst quadrants and value matrices, plus independent case studies (e.g., Reddy Ice), provides credible external validation of business value and deployment at scale.

Technically, Atlas is clearly more than a simple CRUD application: it integrates data from multiple systems, supports multi-process planning (demand, S&OP/IBP, inventory, supply, schedule), runs scenarios, and incorporates machine learning models for forecasting and optimization. Public materials indicate support for multi-echelon inventory optimization, ensemble forecasting and multi-objective optimization, and recent enhancements use generative AI for explainability, which is a meaningful usability improvement in a domain where black-box behavior can hinder adoption.⁹¹¹²⁰ However, the underlying algorithms and data structures remain largely undocumented in public; there is no open technical guide comparable to, for example, Lokad’s detailed descriptions of probabilistic forecasting and stochastic optimization. As a result, claims about “probabilistic planning,” “AI-driven decisions” and “simultaneous multi-objective optimization” should be interpreted as high-level capabilities rather than as evidence of cutting-edge algorithmic innovation per se.

Relative to the broader state of the art in supply chain analytics, JGS appears to occupy a pragmatic position: a capable, cloud-based planning suite that has adopted contemporary AI/ML and optimization concepts and wrapped them in a planner-friendly UI, but without publicly exposing enough technical detail to conclusively assess whether its internals are at the frontier of probabilistic modeling or large-scale stochastic optimization. For many organizations, the combination of functional coverage, usability, deployment track record and vendor support will matter more than the exact mathematical form of Atlas’ models, and on those dimensions JGS has credible evidence of success. For buyers whose primary concern is maximum technical transparency and the ability to inspect or extend the underlying algorithms, however, the lack of detailed public documentation means that vendor-led evaluations, proofs-of-concept and direct technical workshops would be essential before drawing firm conclusions about how “state-of-the-art” the platform is in practice.